KrigR —A Tool for Downloading and Statistically
Downscaling Climate Reanalysis Data.
Efficient Data Retrieval and Processing of ECMWF C3S Products for Your Research
KrigR
GitHub
Find me around the venue if
I’m not here. I’d love to chat.
Erik Kusch & Richard Davy
Kriging
Climate Variable
Reanalysis Data Product
Time-Window
Temporal Resolution
Geographical Region
Directory for NETCDF
Name for NETCDF
CDS API Credentials
Climate Data Retrieval
Covariates
Extent controls spatial limits of the data.
All subsequent steps in KrigR can
handle third-party data.
Rectangular Extent Shapefile Point-Coordinates
Interpolation uncertainty is
an output unique to Kriging.
Computational costs varies
according to specifications.
High uncertainty around
coastlines and islands
ONGOING DEVELOPMENT
1. Widened support for ECMWF CDS data
2. Removal of deprecated dependencies
3. Tiled Kriging algorithm
4. Pre-made data product hosting
THE KrigR-TOOLBOX TO-DATE
R-internal functionality for retrieval of data matching requirements:
1. Downloading & handling of ERA5(-Land) products
2. Provision and preparation of interpolation covariates
3. Statistical interpolation of climate data via Kriging
TResolution and TStep control temporal resolution of the data.
FUN controls aggregate metrics.
The KrigR-workflow combined with CMIP6 data allows creation of high-resolution, bias-corrected projections.
CLIMATE DATA NEEDS OF THE 21ST CENTURY
1. Data Accuracy
Global legacy climate data sets
(e.g., CRU, WorldClim) offer
subpar accuracy
Macroecology relies on
climate data at global scales
Climate Reanalyses (e.g.; ERA5/ERA5-Land) offer higher
accuracy and data uncertainty flags.
Environments
…
3. Range of Variables
Legacy data follows a temperature ( )
–precipitation ( ) paradigm
Neglecting other essential climate
variables (ECVs) like wind ( ),
radiation ( ), etc. (…)
ERA5(-Land) offers up to 83 ECVs.
2. Temporal Resolution
Legacy data sets rarely report
data at sub-monthly intervals
Biological processes ( ) and
extreme events ( ) operate at
finer temporal resolutions
ERA5(-Land) offer data at hourly intervals.
© Connor Bernard
THE KrigR WORKFLOW –3 STEPS TOWARDS PURPOSE-MADE CLIMATE PRODUCTS
Data to downscale
Target resolution or a raster
object whose resolution to match
Optional, shapefiles or points
like specified in download_ERA()
Whether to keep the GMTED
2010 data set on your hard drive
A DEM won’t be suitable for all ECVs.
IMPLICATIONS OF KrigR-DERIVED PRODUCTS
Kriging Uncertainty
1. Statistical interpolation uncertainty ()
constant across temporal scales.
2. Dynamical uncertainty () diminishes
as time-scales increase.
3. Both sources of uncertainty are
important and should be propagated into
downstream analyses.
KrigR Products vs. Legacy Data
1. KrigR-products do not align
with most legacy products.
2. Particularly, in topographically heterogenous regions, KrigR
seems most reliable and informative to us.
Kriging Accuracy
1. Difference of upscaled & interpolated product ()
2. Total uncertainty of kriged product:
1. Kriging outperforms most other methods.
2. Kriging is highly accurate for a variety of ECVs.
Air Temperature Soil Moisture
ROADBLOCKS FOR CLIMATE REANALYSES
1. Accessibility
Climate Data Store (CDS) interface can be overwhelming and
downloads hard to reproduce
CDS APIs (e.g., ecmfr) download specification can be unintuitive
and don’t offer data manipulation
2. Spatial Resolution
Native spatial resolution of climate reanalyses is coarser than
that of legacy data
Practitioners have become accustomed to these fine spatial
resolutions
Practitioners need an intuitive, reproducible R-interface for
data retrieval and handling.
Practitioners require a workflow for creation of high-
spatial-resolution data products.
Localisation of kriging
Date storage directory
Data to downscale
Training covariates
Target Covariates
Parallelisation
Name for file output
Whether to delete
the temporary files
(1 per layer)
